The winning line

About three quarters of the 96.50% lives in the free spins round. That is where the multiplier dice keep adding to one running total instead of wiping it after every spin. Honestly, turnover per round is the only lever you can actually pull here: 448 units grinding, 280 with the ante or 100 buying. Bank any round clearing 100× total bet.

53 keys out of 53, and the page that already holds the tactics

Start with the bit no review site tells a punter. The init payload this game fires off at load is the same as the one Gates of Olympus 1000 sends, matching on every one of 53 keys bar the session timestamp. Same paytable and the same eight reel sets. The scatter table matches too, along with the 16 step multiplier ladder, bet levels and coin values. Buy configuration, win cap fields and the RTP block all line up as well. I reproduced it across two Dice sessions, plus one parent session on the same host that day.

So, really, there is no fresh maths to learn here. Every bankroll gate and trigger calculation carries straight across, and every exit number does too, right to the last decimal onGates of Olympus 1000 . That page already has the base tactics in full. Read them there. No point reading the same thing twice.

What that page cannot cover is the wrapper, and that is where this build can actually sting people. Look, the paytable has had its colour ranking shuffled against the original. The win cap prints two different numbers depending on a toggle, while the ante is unusually cheap in RTP terms but kills the buy button completely.

The whole thing exists because of a licence class. Belgium’s B+ class covers online gaming arcades, and those venues cannot carry classic slot machines. They can carry certified dice games instead, a separate category defined by symbols shown as dice faces. So suppliers repaint the hits with pips and ship the result as a legally different product. Pragmatic moved into that niche in October 2024.

Here is the misconception worth killing straight away. Belgian trade coverage pushes dice games as having some sort of placement or strategy layer. Well, this build has none. There is no roll and no placement. You will not find a pick screen or anything interactive anywhere across seven rules pages either.

Hold that next to the rest of the family and, frankly, the difference gets pretty obvious.Gates of Olympus runs a 500× orb ceiling with a 5,000× cap. Its max win odds are 1 in 697,350, so that one really is running different maths. Gates of Olympus Super Scatter caps at 50,000× and charges 50% for the ante. Mind you, that ante actually lifts RTP to 96.55%, and the game also sells two buy prices. Kingdom of Atlantis Jackpot Play is a reskin as well, but it chopped RTP to 94.50% on the way through. That is a maths downgrade dressed up in a costume. This one, to be fair, is just the costume.

Cube, pips and colour: re-ranking a paytable your reflexes will misread

The nine paying symbols did not just get a fresh coat of paint. They were shuffled into a new order, and honestly, the ranking now has bugger all in common with what a Gates regular has drilled into muscle memory.

Pulled straight from rules page 1 of 7 at an A$2.00 total bet, then checked against the payload’s 13 row paytable array. They match exactly. Three easy reading traps come out of that.

Red used to mean money. In Gates 1000, the red and gold crown tops the paytable at 50×. Here, mind you, red has been knocked down to fourth at 12× while blue takes top spot. Read a screen of red dice like a crown screen and you are off by 4.17 to one. At an A$2.00 total bet, that means A$24 lands where your reflex was expecting A$100.

The plain dice climb with pip count, right up until the logic snaps. Five pips beats four, then three and two follow below it. Fair enough. Then the gold single pip die jumps over the lot at 10×. One pip beating five pips is probably the least readable bit of the whole grid.

Colour is doing two jobs here. The multiplier symbols are dodecahedral gem dice, with green and blue versions plus pink and crowned purple. That means blue and purple can each be either a premium or a multiplier. Look, shape has to come first: a six faced cube pays as a normal symbol, while the many faceted gem die is a multiplier.

So the working read is dead simple. First ask cube or gem die. Then check whether the pips are lightning or plain. Only after that do you read colour: blue first, then green, followed by purple and red.

12,000× and 15,000× are the same A$45,000

Flip the ante and the cap sentence on rules page 6 of 7 changes with it. With ante off, it reads “The maximum win amount is limited to 15000X bet.” Turn ante on and, sure enough, the same sentence on the same page reads 12000X. Everything else stays put.

Nothing else has to move, of course. The multiple is being measured against a total bet that just went up 25%.

The maths is one line: 12,000 × 1.25 = 15,000. The ante bumps the denominator by a quarter, so the multiple drops by a fifth. The cash ceiling lands in exactly the same place (DERIVED from the live bet menu and the cap wording). Anyone treating that toggle as 3,000× of lost upside has been caught by a unit change. Frankly, no source we could find explains it.

The ceiling is mostly decorative anyway. At 1 in 2,749,771 spins, one expected max win chews through A$549,954 of turnover at the A$0.20 floor, or A$5,499,542 at the A$2.00 default. Nobody sensible is shaping a session around that number. The cap wording only matters because, once hit, everything still left in the round is forfeited.

The ante and the buy button cannot both be on

Rules page 3 of 7 stuffs both features into one paragraph, and that is where the exclusivity gets buried. Bet multiplier 25× doubles the natural free spins chance and puts more scatters on the reels. It also kills the purchase feature. Bet multiplier 20×, meanwhile, is the state where the buy is available at 100× total bet.

Take the ante for what it is. Sure, it costs 25% of stake, but the natural trigger rate doubles and the return stays at 96.50%, exactly where the base game sits. That is unusual. Most antes make you pay for the extra speed through a worse edge. This one does not.

That flat RTP matters. A 25% stake bump for double the trigger rate works out to 1.6 times as many rounds per dollar wagered, while the return holds at 96.50%. So the extra feature traffic has to be funded somewhere else in the maths. The base game is the obvious suspect because more scatters on the strips means fewer spots for paying symbols. Honestly, the payload does not break that trade off down, so I am not pretending it does.

The buy is the other route. It costs 100× current total bet, which means A$200 at the A$2.00 default or A$20 at the A$0.20 floor. Four scatters are guaranteed on the trigger spin. Mind you, four is the cheapest trigger at 3× total bet. Five pays 5×, while six jumps all the way to 100×.

That is where the buy really stings, and RTP does not show it. A natural six scatter trigger pays 100× total bet before the first free spin even starts, so the cost of a same stake buy is already covered. Make buying the round a habit and that result simply cannot happen. Run the ante instead and it stays alive, with extra scatters now sitting on the reels.

Four calls, and none of them involve rolling anything

Step 1 — Recode the paytable before the first A$0.20 leaves the wallet

One screen settles the two things worth checking before cash goes anywhere. The recode work is there too. Simple.

Action: Open the info button and go to page 4 of 7. Read the RTP figures properly. They should show 96.50, 96.50 and 96.49. Then go back to 1 of 7 and spend a minute on the paytable. Look, say the order once so it sticks: blue, then green. Purple and red follow, then gold single pip. After that come the white dice from five down to two.

Watch for: The top row is a blue die. If your eye still wants to put red there, well, the recode has not landed yet.

Decision branches: If page 4 of 7 shows 96.50 / 96.50 / 96.49, you are on the certified top tier and Step 2 is live. If any of the three figures is lower, the operator has loaded a reduced tier. No source we found proves reduced Dice tiers are actually out there, so treat a lower figure as a live finding rather than recycled gossip. Take the session somewhere else. A drop to 94.50% burns an extra A$40 per 1,000 spins at A$2.00. If the buy panel is greyed out before you have touched anything, the ante is already switched on and that is what killed the buy. Turn it off. Then read the cap sentence again on page 6 of 7 and watch the number shift from 12000X back to 15000X.

Step 2 — Fund the 448-spin wait, because the base game never compounds

The two sides of this pokie play by different rules, and page 2 of 7 spells it out. In the base game, the multiplier dice left on screen when a tumble sequence ends are added together. That total hits the sequence, then disappears. Free spins work differently. Every die landing on a winning spin feeds a running total, and that total keeps applying to every winning spin still to come in the round.

So the base game is a one spin lottery. No memory. Nothing to chase. The round is the only part that actually accumulates anything, so, to be fair, either fund the wait properly or the whole plan is just decoration.

Action: Set the coin value so the stash can wear 900 base spins. That is roughly two expected waits at 1 in 448 and gives about an 87% chance of seeing at least one natural trigger (DERIVED, geometric). You need A$180 at the A$0.20 floor or A$1,800 at A$2.00. If the budget cannot carry that, frankly, cut the coin value first. Do not gut the run length.

Watch for: the coin value box is what matters here, not the total bet. Total bet is locked at coin value times 20. So, honestly, bumping the coin from A$0.01 to A$0.02 doubles what each spin costs and chops the funded run in half with one click.

Decision branches: If the stash will not fund 450 spins → drop to the A$0.20 floor. That buys 900 spins for A$180, otherwise go grind a lower variance pokie because the wait is more likely to kill the session than the round is to arrive. If a big multiplier die drops in the base game and the sequence dies without paying → nothing vanished and the game owes nothing. Dead spin dice get binned in both modes. Treating one as a warning that a hit is coming is the classic near miss error from Tversky and Kahneman, 1974. If the trigger lands → judge the round by when the chunky ladder steps show up, not by how many spins have gone. Only 4 of the 16 ladder values are 100× or higher. If the total is still below 10× at free spin 8 of 15, one of those four has to land just to clear the buy price. Mind you, the weightings are unpublished, so that is simple counting, not a probability.

Step 3 — Price the ante at 25% against half the wait

The ante is a turnover tool, nothing more. It does not move the edge. The only useful question is whether the shorter road to a round is worth the bigger bill getting there.

Action: with the ante switched on, total bet jumps to 25 coins instead of 20. That means A$0.25 at the floor coin and A$2.50 at the default. Well, budget 450 ante spins to get the same 87% confidence the grind lane needed 900 for. The bill is A$112.50 at the floor or A$1,125 at A$2.50. That is 37.5% less cash for the same odds of seeing a round, DERIVED, and it still leans on the unverified 1 in 448 base figure.

Watch for: the buy panel goes grey as soon as the toggle flips. The cap sentence on page 6 of 7 also changes to 12000X. Both are meant to happen. Neither means the game has been downgraded.

Decision branches: If the operator blocks the buy through bonus terms or simply removes the panel → run the ante. It is the only accelerator left on screen and, to be fair, at 96.50% it does not cost extra edge. If the stash cannot cover ten buys at the chosen stake → use the ante instead. A two buy budget is just a coin toss pretending to be a plan. If the bankroll covers ten buys and the buy is live → leave the ante off and move to Step 4 using the arithmetic in the maths block below. If the stake starts creeping toward A$300 because the ante ladder allows it → that is the ante blowing out the top of the range, not the game lifting its ceiling. At A$300 a spin, the 450 spin ante gate costs A$135,000. That is not a plan.

Step 4 — Buy at 100× only with ten buys in the stash

The buy costs 100× total bet and guarantees a four scatter entry at 96.49%. It is the lowest turnover route into the only part of this game that compounds. Look, it is also the quickest way to skin a wallet because 448 spins of variance get crushed into one button press.

Action: set the series before buying the first round. Ten buys minimum. That is 1,000× total bet, or A$200 at the A$0.20 stake and A$2,000 at A$2.00. Bank it and shut the session when any round clears 100× total bet. Sure, 100× is the game’s own printed price for one round, so there is no need to invent a fancier exit.

Watch for: watch the running multiplier total through every bought round. Also watch for three scatters landing inside it. Three is enough to retrigger for 5 more spins even though four were needed to enter, and the rules print no retrigger cap.

Decision branches: If a round clears 100× total bet → close it out. There is no press mechanic or gamble ladder here, and this build has no second tier of buy waiting to turn a good round into something better. If ten straight buys each return under 30× → that is 1,000× staked for under 300× back. Frankly, the documented failure shape on this engine is exactly that pattern stretched out. On the parent title, an AskGamblers thread records a punter taking 150 plus consecutive 100× buys with a best result near 300×. Each buy is independent. Buy twenty one knows nothing about the twenty before it. If the buy panel will not respond → check the ante toggle before blaming the casino, because the two modes are mutually exclusive by design. If the urge is to raise coin value and win back a bad series → do not. That shortens the series exactly when it needs length, while the ante has already lifted cost by 25% before any manual escalation gets involved.

Turnover per round is the whole comparison

Because all three lanes return within one basis point of each other, the edge stops deciding the expected cost. What matters is how much money has to pass through the machine before one round turns up. Honestly, that is the whole decision. It fits in a table.

Look, the buy is 2.8 times leaner than the ante lane and 4.5 times leaner than grinding because it simply shoves less money through the machine (DERIVED; the 448 base figure is UNVERIFIED and the 224 comes from the rules’ own word “doubled”). The trade is ugly but simple. You get the worst trigger of the lot.

Where the return sits tells you why the wait matters so much. Roughly 20 to 30 points of the 96.50% live in the base game, while free spins carry 70 to 80 points (DERIVED from the orb reset asymmetry on rules pages 2 and 3 of 7). Honestly, a punter who quits after 300 spins has, in expectation, only played the thin end of the game. The base alone gives back nowhere near the number splashed across the info screen.

One more number drops straight out of the buy price. A 100× purchase returning 96.49% already includes the 3× scatter pay on the trigger spin, so the round itself averages about 93.5× total bet (DERIVED). At A$2.00, that works out to A$187 back from a A$200 button press. Mind you, the average hides a savagely crooked distribution, so ten buys is the floor for testing this lane, not some magic target.

There is no PAR sheet for this build, and none for the parent either. Against the UKGC 2024 band of 8 to 20 for very high variance titles, the spec points to σ around 12 to 15 per unit staked (ESTIMATED). At that sort of variance, 900 spins at A$2.00 carries an expected loss near A$63, with several hundred either side well inside the swing. That is why the gates above are written around spins funded and buys funded. A stop loss percentage tells you bugger all here.

Community data is useful for the maths. That is about it. Slot Tracker logged 1,421,870 spins on the Gates engine at a tracked 84.3%, with heavy selection bias, while dry runs of 600 to 800 base spins between triggers show up in the parent title’s records. Sure, those belong to the parent build and are quoted that way. Nobody has posted a session on this one, and the variable ratio reinforcement behind the chasing (Schull, 2012) works the same no matter what artwork is slapped on the reels.

Where a costume costs money

1. Reading 12,000× as a cut ceiling. The cash number is A$45,000 either way. This is the most misread thing on the build and every source we checked leaves it unaddressed.
2. Pricing a screen by colour. Red is fourth here at 12× and blue is first at 50×. Gates muscle memory misvalues a full red screen by 4.17 to one.
3. Hunting the dice decision. There isn’t one. Seven rules pages, no roll, no placement, no pick. The dice are art direction and the strategy layer belongs to a category this game only borrows the label from.
4. Serial buying after a dry run. Twenty bad buys change nothing about the twenty-first. The ten-buy series exists so the decision to stop is made before the losses argue for continuing.
5. Toggling the ante and then waiting for the buy button. They are mutually exclusive. The panel greys out on purpose and it is not a fault.
6. Chasing the printed A$300. That figure needs 25 coins at A$12.00, meaning the ante is on, meaning the buy is dead and the cap prints 12,000×. The ante-off ceiling is A$240.

Stop numbers for a game you can already play under its own name

Funding gate, grind lane: You want 900 base spins funded at the chosen stake. That means A$180 at the A$0.20 floor, or A$1,800 at A$2.00. Well, under 450 funded spins, the natural trigger probably never turns up.

Funding gate, ante lane: Fund 450 spins at 25 coins. That is A$112.50 at the floor coin and A$1,125 at A$2.50. Shorter than that and you are barely giving the lane a run.

Funding gate, buy lane: Ten buys means 1,000× total bet. That is A$200 at the floor or A$2,000 at A$2.00. To be fair, anything below that is not really opening the lane in the first place.

Bank line: Any round clearing 100× total bet has done its job because that is the game’s own price for one round. Take it. Nothing in this build compounds after the round is finished.

Buy-series stop: Ten straight buys under 30× each is enough. Frankly, close the lane there. Do not hang around until the losses start feeling like they are due to turn.

Recode check: If red still sits as the top symbol at spin 50, the paytable has not arrived yet. Drop straight back to the floor stake until it does. No heroics.

Escalation line: the coin value picked at Step 2 stays locked for the whole session. Every gate above assumes that exact value, and bumping it up halfway through blows the lot.

One note on where this turns up. Pragmatic Play holds no Interactive Gambling Act 2001 licence for Australia, so an Australian only gets to this title through offshore licensed operators. The demo gateway serves it worldwide in AUD with no geo block observed. One offshore crypto operator was directly confirmed to carry it. Two others show up in search indexes, but could not be verified in this research pass, so don’t tell punters they stock it. Mind you, Belgian arcade sites will not take an Australian account anyway. Their stake setup is still useful as a comparison: a B+3858 licensee caps the top total bet at EUR 1.60 against the game’s own printed ceiling, an operator clamp of roughly 187 to one. That is arcade house policy, not statute. The Belgian Gaming Commission publishes hourly loss ceilings of EUR 12.50, EUR 25 and EUR 70 for land based venues. It also states that no such figure is yet set for online play.

And the honest recommendation. Honestly, there is no reason for an Australian to hunt down this build. The maths is byte for byte the parent’s. Community records and the guides point punters back to the original, while the artwork is the only real change. If this is the version sitting in the lobby, fine, it is the same pokie and the numbers above still apply. If there is a choice, take Gates of Olympus 1000 for the ecosystem rather than the maths, or Gates of Olympus if a 1-in-697,350 ceiling fits a smaller bankroll better than a 1-in-2.75-million one. Sure, other Pragmatic Play slots stretch the ceiling further, but none of them turn this into anything different from its parent.

Strategy by Dave Holloway, Floor Veteran. Last updated: 19 Aug 2026.
18+. Gamble responsibly. Gambling Help Online 1800 858 858 | Lifeline 13 11 14.